Using annual Landsat imagery to identify harvesting over a range of intensities for non-industrial family forests

摘要: Abstract The monitoring of forested landscapes dominated by many small private forest owners is difficult or not possible without spatially explicit and up-to-date information on land cover change. Analysis time series multispectral data from the Landsat satellites have spatial temporal characteristics required to detect sub-hectare non-stand replacing harvest events over large areas. We identified harvests that occurred in six western upper Michigan counties 1985 2011 using best available pixel (BAP) image composites Composite2Change (C2C) approach. detected a total 7071 harvesting with size ranging 0.5 171.36 ha average 6.42 ha, analyzed their trajectory. To gain confidence our mapping, we compared findings overlapping decade Global Forest Watch (GFW) data. Agreement between datasets was high, 94.24% C2C GFW pixels same change year improving 98.74% within ±1 year. This automated detection system, which can capture otherwise missed harvests, valuable natural resource agencies responsible for compliance regulations areas, researchers requiring estimates levels nature status trends family forests.